p53 protein aggregation promotes platinum resistance in ......ORIGINAL ARTICLE p53 protein aggregation promotes platinum resistance in ovarian cancer Y Yang-Hartwich 1, MG Soteras

ORIGINAL ARTICLE

p53 protein aggregation promotes platinum resistancein ovarian cancerY Yang-Hartwich1, MG Soteras1, ZP Lin2, J Holmberg1, N Sumi1, V Craveiro1, M Liang1, E Romanoff1, J Bingham1, F Garofalo1,A Alvero1 and G Mor1

High-grade serous ovarian carcinoma (HGSOC), the most lethal gynecological cancer, often leads to chemoresistant diseases. Thep53 protein is a key transcriptional factor regulating cellular homeostasis. A majority of HGSOCs have inactive p53 because ofgenetic mutations. However, genetic mutation is not the only cause of p53 inactivation. The aggregation of p53 protein has beendiscovered in different types of cancers and may be responsible for impairing the normal transcriptional activation and pro-apoptotic functions of p53. We demonstrated that in a unique population of HGSOC cancer cells with cancer stem cell properties,p53 protein aggregation is associated with p53 inactivation and platinum resistance. When these cancer stem cells differentiatedinto their chemosensitive progeny, they lost tumor-initiating capacity and p53 aggregates. In addition to the association of p53aggregation and chemoresistance in HGSOC cells, we further demonstrated that the overexpression of a p53-positive regulator,p14ARF, inhibited MDM2-mediated p53 degradation and led to the imbalance of p53 turnover that promoted the formation of p53aggregates. With in vitro and in vivo models, we demonstrated that the inhibition of p14ARF could suppress p53 aggregation andsensitize cancer cells to platinum treatment. Moreover, by two-dimensional gel electrophoresis and mass spectrometry wediscovered that the aggregated p53 may function uniquely by interacting with proteins that are critical for cancer cell survival andtumor progression. Our findings help us understand the poor chemoresponse of a subset of HGSOC patients and suggest p53aggregation as a new marker for chemoresistance. Our findings also suggest that inhibiting p53 aggregation can reactivate p53pro-apoptotic function. Therefore, p53 aggregation is a potential therapeutic target for reversing chemoresistance. This isparamount for improving ovarian cancer patients’ responses to chemotherapy, and thus increasing their survival rate.

Oncogene advance online publication, 29 September 2014; doi:10.1038/onc.2014.296

INTRODUCTIONOvarian cancer is the most lethal gynecological malignancy.Although most patients initially respond to chemotherapy, themajority succumb to recurrent chemoresistant tumors.1 Efforts toovercome chemoresistance have been largely unsuccessful. Themortality rate remains high. High-grade serous ovarian carcinoma(HGSOC) accounts for 67% of all ovarian cancers and is the mostaggressive subtype. A main characteristic of HGSOC is that 96% ofthe tumors bear p53 mutation.2 p53 is a central transcriptionalmediator. By binding to DNA, p53 controls the expression ofhundreds of target genes in order to maintain homeostasis andgenome integrity. In ∼ 90% of ovarian cancers, one of the p53alleles has a mutation that abrogates p53 transcriptional activity.3,4

The other wild-type allele is usually also inactive or attenuated.5

On the other hand, many tumors without p53 mutation still harbora transcriptionally inactive form of p53,6–8 suggesting that geneticmutation is not the only cause of p53 inactivation. Therefore, thepresence of wild-type (wt) p53 does not necessarily indicate goodprognosis and chemoresponse. In the case of HGSOC, patientswith wt p53 have even poorer survival rates and are morechemoresistant than those with mutant p53.9 Therefore, addi-tional mechanisms of p53 inactivation beyond genetic mutationsneed to be further studied.

Recently, it has been shown that in different types of cancer celllines and tumors, both mutant and wt p53 protein can aggregateinto amyloid fibrils.10–13 Protein aggregation is a pathogenicfeature of a growing number of diseases, such as Alzheimer’s andParkinson’s disease.14 In vitro studies suggest that p53 is anamyloid-forming protein. Amyloid-forming proteins are an unu-sual subset of proteins that are able to aggregate. Although theamino-acid sequences of these proteins are diverse, they all adopta similar, highly organized structure upon aggregation known asthe cross-β spine that consists of an ordered arrangement ofβ-sheets. The transactivation, DNA-binding and tetramerizationdomains of p53 protein can all misfold and form amyloid fibrillaraggregates.15–19 Protein aggregation can perturb essential cellularfunctions and cause various human disorders.20 Considering thecrucial role of p53 in maintaining cellular homeostasis, theaggregation of p53 may affect its normal functions and act as akey factor in the initiation and progression of cancer.21,22

Cancer stem cells (CSCs) are a subpopulation of chemoresistanttumor cells that can give rise to heterogeneous tumors throughself-renewal and multilineage differentiation. CSCs were shown toexist in hematologic and solid cancers23–25 including ovariancancer.26–28 The CSC model explains much of ovarian cancer’setiology, such as tumor dormancy, minimal residual disease, drug

1Department of Obstetrics and Gynecology, Yale University School of Medicine, New Haven, CT, USA and 2Department of Pharmacology, Yale University School of Medicine,New Haven, CT, USA. Correspondence: Professor G Mor, Division of Reproductive Sciences, Department of Obstetrics, Gynecology and Reproductive Sciences, Yale UniversitySchool of Medicine, 333 Cedar Street, LSOG 305A, New Haven, CT 06520, USA.E-mail: [email protected] 19 March 2014; revised 25 June 2014; accepted 31 July 2014

Oncogene (2014), 1–12© 2014 Macmillan Publishers Limited All rights reserved 0950-9232/14

www.nature.com/onc

resistance and disease relapse. Ovarian cancer stem cells (OCSCs)have been identified with different markers. We identified apopulation of CD44+/MyD88+ OCSCs that can initiate tumors inimmunocompromised mice and differentiate into multiple celltypes including CD44− /MyD88− ovarian cancer cells (OCCs).29–31

More important, CD44+/MyD88+ OCSCs are in vivo and in vitroresistant to currently available chemotherapeutic agents, includ-ing taxane- and platinum-based agents.29,32 Among the CD44+/MyD88+ OCSCs isolated from 4100 ovarian cancer samples, wehave identified two cell lines with wt p53. They are highly resistantto platinum-based agents and have been used as a model ofOCSCs with p53 aggregates in our study. We hypothesize that theaggregation of p53 can inhibit its pro-apoptotic functions andpromote chemoresistance in OCSCs.The objectives of this study are twofold. First, we determine

whether p53 aggregation is associated with the loss of p53 pro-apoptotic functions and chemoresistance. Second, we explore thepotential molecular mechanisms causing p53 aggregation. It ishypothesized that the formation of protein aggregates dependson protein concentration, complex interactions with otherproteins and specific cellular environment.33 However, the causeof p53 aggregation in tumors is still unclear. We propose that

inefficient protein degradation leads to the accumulation ofmisfolded p53 proteins and induces aggregation.Using the OCSC model, we have uncovered that MDM2 is

inhibited by an antagonist p14ARF (ARF), and the lack of MDM2-mediated degradation can cause p53 to accumulate and formaggregates. ARF is a tumor-suppressor protein that binds toMDM2 and inhibits the E3 ligase activity of MDM2, leading to p53protein stabilization. ARF also promotes MDM2 degradation andsequesters MDM2 into the nucleolus.34–37 It is well known thatMDM2/MDM4 overexpression causes p53 inhibition.38 The ampli-fication of MDM2/MDM4 is frequently observed in tumors.39 Ourin vitro and in vivo data reveal a paradoxical mechanism by whichthe inhibition of MDM2 by ARF can result in p53 inactivation andchemoresistance by causing the imbalance of p53 turnover, andp53 protein aggregation.

RESULTSCD44+ OCSCs, but not their CD44− progeny OCCs, possesstumorigenicity and chemoresistanceUtilizing the previously reported method,29,30 we isolated apopulation of CD44+/Myd88+ OCSCs from HGSOC tumors. CD44

Figure 1. CD44+ OCSC1s but not their progeny CD44− OCC1s possess cancer stem cell properties. (a) Schematic representation of theisolation of OCSCs and OCCs. CD44+ OCSCs are isolated from HGSOC tumors or ascites by flow sorting and cultured. They were infected bylentivirus expressing GFP and then subcutaneously injected to immunocompromised mice. From the formed tumors, GFP+/CD44− OCCs areisolated and cultured. Fluorescence-activated cell sorting (FACS) analysis (dot plot) indicates that GFP+/CD44+ OCSCs can in vivo differentiateinto GFP+/CD44− OCSCs. (b) Morphology changes of OCSC1s and OCC1s in carboplatin treatment. (c) Dose-dependent viability of OCSC1s(red) and OCC1s (blue) treated with carboplatin. (d) Caspase-3 (left) and caspase-9 (right) activities of nontreated (dotted) and carboplatin-treated (solid) OCSC1s (red) and OCC1s (blue).

p53 aggregation and chemoresistanceY Yang-Hartwich et al

2

Oncogene (2014), 1 – 12 © 2014 Macmillan Publishers Limited

+/Myd88+ OCSCs possess cancer stem cell properties includingtumor-initiating capability (Supplementary Figures 1a–c). Todemonstrate that these OCSCs have tumor formation anddifferentiation capacity, we labeled them with green fluorescentprotein (GFP)-expressing lentivirus and subcutaneously injectedthem into immunocompromised mice (Figure 1a). CD44+/GFP+OCSCs formed heterogeneous tumors consisting of both CD44+/GFP+ and CD44− /GFP+ cells. The CD44− /GFP+ cells isolatedfrom these tumors lost cancer stem cell properties and cannotform tumors in mice (Figure 1a and Supplementary Figures 1a–dand 2a). We named them CD44− OCCs.In this study, we selected two OCSC clones with wild-type p53

that were isolated from two individual patients. The OCC lineswere derived from both OCSC clones. These two sets of cell lineswere named OCSC1/OCC1 and OCSC2/OCC2 (SupplementaryFigures 1 and 2a, respectively). The p53 mutation analysis wasperformed on these cells by sequencing full-length p53 com-plementary DNA (cDNA) and exons 5–8 of TP53 gene (encodingthe DNA-binding domain of p53 protein that is frequentlymutated). We did not detect p53 mutations in OCSC1, OCC1,OCSC2 or OCC2 (Supplementary Tables 1 and 2). We did notdetect the overexpression of dominant-negative p53 isoforms orΔN-p63 (data not shown).OCSCs are chemoresistant. When they differentiated into OCCs,

they become chemosensitive. For example, when treated withcarboplatin (50 μg/ml for 48 h), OCSC1s maintained their normalmorphology. The same treatment induced significant cell death inOCC1s (Figures 1b and c). We detected the induction of caspase-3and -9 activities after carboplatin treatment in OCC1 and OCC2,but not in OCSC1 and OCSC2 (Figure 1d and SupplementaryFigures 2b and c). Together, our data demonstrate that OCSCs, butnot their progeny OCCs, possess tumor-initiating properties andresistance to platinum-based treatment.

The p53 protein aggregation is associated with p53 inactivation inOCSCsBy immunofluorescent staining, we detected that OCSCsexpressed high levels of p53 (Figure 2a and SupplementaryFigure 3d). At the same time, we observed the presence ofaggregates in OCSCs. We performed immunofluorescent stainingwith thioflavin T, a fluorescent dye that labels amyloidaggregates40) and anti-amyloid fibril antibody (OC, an antibodyrecognizing generic epitopes common to amyloid fibrils andoligomers41). OCSCs and OCCs were co-stained for aggregates andp53 (Figures 2a and b and Supplementary Figures 3a and b).OCSCs showed widespread protein aggregation that colocalizedwith p53 staining in both the control and carboplatin-treatedgroups. The positive staining localized in both nuclei andcytoplasm. However, it mainly concentrated in the nuclei(Supplementary Figure 4). OCCs were negative for proteinaggregation, even when p53 protein accumulated followingcarboplatin treatment. When we knocked down the expressionof p53 in OCSC1 with small hairpin RNA (shRNA), the levels ofaggregates detected by staining were significantly decreased(Supplementary Figure 5). Our observation differed from theprevious report that showed cytoplasmic distribution of p53aggregates.11 In the nuclei of OCSCs, the levels of aggregates aremuch higher than in their cytoplasm that was in a similar patternas p53 levels. Our observation indicates that in OCSCs most of thep53 proteins were translocated into the nucleous and formaggregates in the nucleous, and this is different from theaggregates from mutant p53 in the previous report.11 The nuclearenvironment may have a role in regulating the formation of theseaggregates.Western blots from non-denaturing gels (native polyacrylamide

gel electrophoresis) confirmed the presence of high-molecular-weight p53 aggregates in OCSC1, but not in OCC1 (Figure 2c).

Anti-amyloid fibril dot blots showed that OCSC1 and OCSC2 werepositive for protein aggregation, whereas OCC1 and OCC2 werenegative (Figure 2d and Supplementary Figure 3c). We hypothe-sized that the aggregation of p53 protein led to p53 inactivation inOCSCs that might contribute to the resistance to chemo-treatments.The inactivation of p53 is associated with the attenuated ability

of p53 to bind to DNA and activate downstream target geneexpression. To test the effect of aggregation on p53 activity, wetested whether the transactivation of wt p53 was inhibited inOCSC1 and OCSC2. The p53 chromatin immunoprecipitation dataindicated that the binding of p53 to the promoter of MDM2, P21and Puma was inhibited in OCSC1s when compared with OCC1s.Carboplatin treatment slightly increased p53 DNA binding inOCSC1, whereas in OCC1s carboplatin significantly increased p53DNA binding (Figure 2e). Quantitative reverse transcriptase–PCR ofp53 target genes, including MDM2, P21, Puma, Bax and Fas,showed consistent result. In OCC1s when p53 DNA binding wasincreased by carboplatin treatment, the target gene transcriptionwas consequently upregulated (Figure 2f). On the contrary, inOCSC1s the inhibited p53-DNA binding led to the inhibitedresponse of p53 target gene transcription. Carboplatin treatmentfailed to upregulate p53 target gene mRNA (Figure 2f). Similarresults were observed in OCSC2s and OCC2s (SupplementaryFigures 3b and d). Our data indicate that the transactivation of wtp53 was inhibited in OCSCs and restored when they differentiatedinto OCCs.In order to determine whether the in vitro observations are also

present in tumor samples, we performed a study that included 32ovarian tumor biopsies (Supplementary Table 3). We detected thepresence of protein aggregates in almost half of the samples.More important, we found a significant correlation between p53overexpression and protein aggregation (Figure 2g andSupplementary Figure 6). Out of 18 patients with high p53 levels,13 patients had protein aggregates. Out of 14 patients with lowp53 levels, 13 patients showed low levels of aggregation. Based onthe amyloid-forming nature of p53 protein and our in vitro data,we conclude that the overexpressed p53 protein is involved in theaggregates, associated with the inhibition of p53 transcriptionalactivity in OCSCs.

ARF overexpression inhibits p53 degradation in OCSCsUsing immunofluorescent staining and western blot, we detectedhigh levels of p53 protein in OCSC1s and OCSC2s regardless ofcarboplatin treatment. Conversely, OCC1s and OCC2s expressedlow basal levels of p53 protein. However, following carboplatintreatment, OCC1s and OCC2s showed significant increase of p53protein level and nuclear translocation (Figures 2a and 3a andSupplementary Figure 3a). OCC1 also showed a significantupregulation of MDM2 in response to carboplatin. Conversely,the MDM2 expression did not change in OCSC1s followingcarboplatin treatment (Figure 3a).MDM2 and MDM4 are the major regulators of p53 protein

degradation. OCCs expressed higher levels of MDM2 and MDM4than OCSCs (Figure 3a and Supplementary Figure 7). However, wedid not observe differences between OCSC1s and OCC1s in MDM2mRNA level (Figure 3b). When OCSC1s were treated by aproteasome inhibitor, MG132, the level of MDM2 protein inOCSCs was significantly upregulated, suggesting that MDM2 israpidly degraded in OCSC1s (Figure 3c). Interestingly, when weoverexpressed MDM2, the degradation of p53 was induced inOCC1s, but not in OCSC1s (Figure 3d). These findings indicate thepresence of factor(s) that can promote MDM2 degradation andinhibit MDM2-mediated p53 degradation, leading to p53 accu-mulation in OCSCs.Tumor suppressor ARF binds to MDM2, inhibits the E3 ligase

activity of MDM2, promotes MDM2 degradation and sequesters


3

© 2014 Macmillan Publishers Limited Oncogene (2014), 1 – 12

MDM2 into the nucleolus. By western blot we detected theoverexpression of ARF in OCSC1s (Figure 3a). To determinewhether there is a correlation between the overexpression of ARF

and p53, we knocked down ARF mRNA and protein expression inOCSC1s using two lentiviral shRNAs: shARF1 and shARF2.Inhibition of ARF in OCSC1 led to MDM2 stabilization and the


4


degradation of p53 (Figure 3e). These data suggest that highlevels of ARF in OCSCs inhibit MDM2 function and lead to p53accumulation.ARF is activated by oncogenic insults. As the tumor progresses,

ARF is often inactivated by deletion, mutation or epigeneticsilencing, such as promoter DNA methylation.42 We hypothesizethat when OCSC1s differentiate into OCC1s, epigenetic silencingcaused the decrease of ARF expression that may indirectly reversep53 aggregation. We next examined ARF promoter status bymethylation-specific PCR. ARF promoter is unmethylated inOCSC1s. In OCC1s, it is partially methylated that is sufficient tosuppress ARF expression (Figure 3f). We evaluated the ARF statusof 17 cancer cell lines derived from ovarian cancer biopsies. Basedon their phenotypes including CD44 expression levels (Figure 3g),growth rate, morphology and chemoresponse (data not shown),the cell lines were categorized as either OCSC type or OCC type.The methylation-specific PCR results showed that in 10 OCSC-typecell lines, ARF promoter was unmethylated, and ARF mRNA levelswere significantly higher, whereas 7 OCC-type cell lines showeddifferent levels of methylation and low levels of ARF mRNA(Figures 3g and h).

ARF silencing reverses p53 aggregation and sensitizes OCSCs tocarboplatin treatmentMisfolded polypeptides and damaged mature proteins aredegraded in cells via ubiquitination or autophagy.43,44 Inefficientproteasome degradation leads to protein aggregation affectingnormal cellular functions. We postulated that the overexpressedARF inhibited MDM2-mediated p53 degradation, resulting in theaccumulation of misfolded p53. The p53 protein formedaggregation that inhibited normal p53 functions in OCSCs (modelshown in Figure 4a, left). We therefore examined whetherinhibiting ARF can decrease p53 aggregation. Indeed, ARFknockdown diminished protein aggregation, and had the similareffect as knocking down p53 with shRNA (shp53, Figure 4b andSupplementary Figure 6). Moreover, inhibiting ARF restores p53function and sensitizes OCSC1s to carboplatin treatment(Figure 4c). Our data suggest that when OCSCs differentiate intoOCCs, ARF gene silencing by promoter methylation indirectlyimpedes p53 aggregation, restoring the function of p53 (modelshown in Figure 4a, right).Using a co-culture model, we tested whether it is possible to

sensitize heterogeneous tumors to carboplatin by inhibiting p53aggregation with ARF knockdown. GFP-labeled OCSC1s andmCherry-labeled OCC1s were mixed and cultured together,mimicking heterogeneous tumors. Upon 48 h of carboplatintreatment followed by 48 h of recovery, the only cells thatsurvived were GFP-OCSC1s. When ARF was inhibited in OCSCs byshARF2, we were able to eliminate all the cancer cells withcarboplatin (Figure 4d).

We then tested whether we could reverse chemoresistanceusing an in vivo xenograft model. The mCherry-labeled controlpLKO-OCSC1s or shARF2-OCSC1s were injected intraperitoneallyinto immumocompromised mice. shARF2 does not significantlyaffect the tumor-forming ability of OCSC1s (SupplementaryFigure 8). During cisplatin treatment, tumors in the control groupshowed significant progression, whereas no significant tumorgrowth was observed in the shARF group (Figures 5a and b). Non-denaturing gel and western blot analysis showed that the high-molecular-weight p53 aggregates were downregulated by shARF2(Figure 5c). In the cisplatin-treated groups, shARF tumors showedsignificantly higher levels of active caspase-3 and -9 whencompared with the control pLKO tumors (Figures 5d and e).Morphologically, cisplatin induced extensive cell death in shARFtumors but did not affect control pLKO tumors (Figure 5f).Collectively, our data suggest that cisplatin resistance of OCSC-formed tumors is associated with high levels of p53 aggregates.ARF knockdown can inhibit p53 aggregation, restore p53 functionand sensitize tumors to cisplatin treatment.

Aggregated p53 gain of functionNext, we determined whether the aggregated p53 (AGp53) gainedthe ability to interact with other proteins and potentiallycontributed to tumor progression. We analyzed p53 proteincomplex that was immunoprecipitated from carboplatin-treatedOCSC1s and OCC1s using two-dimensional gel electrophoresis(Supplementary Figure 9). Our analysis indicated a compellingdifference in the profile of p53-binding proteins in OCSC1 withAGp53 and OCC1 with normal wt p53. Using mass spectrometry,we identified a list of proteins that were significantly enriched inthe p53 protein complex of OCSC1 (Table 1). These proteins areinvolved in multiple essential cellular functions, includingcytoskeletal functions, RNA processing, transcription and transla-tion, indicating that p53 aggregation may facilitate tumorprogression through affecting these functions.

DISCUSSIONWe demonstrate that in a population of CD44+ OCSCs, defectiveMDM2-mediated degradation causes p53 protein aggregation.The aggregation sequestrates native p53 protein into an inactiveconformation lacking pro-apoptotic functions that correlates withresistance to carboplatin treatment. Our data reveal a morecomplex scenario of the p53 status and emphasize that in additionto genetic mutations, the imbalanced proteostasis can alsoinactivate p53. Notably, p53 aggregation is cell-type specific andassociated with cancer cells with cancer stem cell properties.Furthermore, we demonstrate that overexpression of ARF may beresponsible for p53 aggregation. The inhibition of ARF can restorep53 degradation and is sufficient to suppress p53 aggregation andreverse chemoresistance.

Figure 2. The p53 protein aggregation is associated with p53 inactivation in OCSC1s. (a) Co-immunofluorescence staining of p53 (red) andamyloid fibrils by Thioflavin T (ThioT, green). Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI, blue). Merged images show theoverlap of p53 and ThioT staining (yellow or orange). Scale bars are 36 μm. (b) Co-immunofluorescence staining of p53 (red) and amyloidfibrils (anti-amyloid fibril antibody, OC, green). Merged images show the overlap of p53 and OC staining (yellow or orange). Scale bars are36 μm. (c) P53 western blot of non-denaturing gel (native polyacrylamide gel electrophoresis (PAGE)) and denaturing gel (SDS–PAGE). In thepresence of aggregates, p53 protein shifted to higher molecular weight. (d) Dot blot of amyloid fibrils in carboplatin treated (+) andnontreated (− ) OCSC1 and OCC1. (e) Chromatin immunoprecipitation (CHIP)–quantitative PCR (qPCR) analysis of p53 DNA binding. Cells werenontreated (NT, black) or carboplatin-treated (Carbo, gray). Primers for qPCR target the p53-binding sites in MDM2, p21 and puma promoter.Data were normalized to nontreated OCSC1s. GAPDH was included as PCR negative control. IP IgG was used as IP negative control. *Po0.05vs NT OCSC1.*Po0.05 vs NT OCC1. (f) Reverse transcriptase (RT)–qPCR of p53 target genes. OCSC1s (blue) and OCC1s (red) were treated bycarboplatin. Data were normalized to the respective nontreated OCSC1s or OCC1s. *Po0.05 vs nontreated control. (g) Correlation analysis ofp53 expression and protein aggregation in tumor biopsies (n= 32). p53 western blot and protein aggregation dot blot were quantified. Thestatistical significance of the correlation between p53 overexpression and protein aggregation is tested by Fisher’s exact test (P-value isindicated in the graph). The above data represent the means± s.d. of three technical replicates.


5


ARF is a key mediator of p53–MDM2 interaction. In addition toshowing the correlation of ARF overexpression and p53 aggrega-tion, we were able to inhibit aggregation and reactivate p53 byknocking down ARF. To our knowledge, this is the first

demonstration of a mechanism causing endogenous p53 aggre-gation in cancer cells. When OCSCs differentiated into OCCs, ARFwas silenced via ARF promoter methylation that potentiallyrestored p53 degradation and inhibited aggregation,


6


consequently reactivating the pro-apoptotic function of p53. Wehave described a unique epigenetically regulated process ofreversing endogenous protein aggregation during cell-typetransition.The formation of high-molecular-mass species and self-

aggregation of p53 was first described in the early 1990s.45 Later,the prion-like characteristics of p53 protein were demonstratedin vitro.15–18 More recent studies have shown the presence of p53aggregates in cancer cell lines and tumor samples, includingbreast cancer, colorectal cancer, skin cancer andneuroblastoma.11,12,22 However, p53 aggregation had not beenstudied in ovarian cancer. Our data suggest that the presence ofprotein aggregates is not rare in ovarian tumors. Proteinaggregation is often associated with p53 overexpression inovarian tumors. The molecular evidences we provide raise thepossibility that p53 protein aggregation is responsible forchemoresistance in a subset of ovarian cancers.In the present study we utilized two chemoresistant ovarian

cancer cell lines with wt p53. We provided evidence supportingthat the overexpressed p53 in OCSCS is indeed wild type, eventhough it lacks pro-apoptotic function. First, no mutations weredetected by DNA or cDNA sequencing. Second, the dominant-negative isoforms of p53 were not detected. Third, once OCSCsdifferentiated into OCCs, protein aggregation diminished and p53pro-apoptotic functions were restored, resulting in OCSCsresponsive to carboplatin.The model with wt p53 aggregation described in this study is

important as it allows us to study the effects of aggregation andgenetic mutation separately. We can study how protein aggrega-tion attenuates p53 activity without the involvement of mutantp53 dominant-negative effects and gain of function. Moreover,our model may explain why HGSOC patients with wt p53 showedpoorer survival rates and were more chemoresistant than thosewith mutant p53.9

The formation of aggregates is not limited to wt p53. Themajority of ovarian tumors have one mutant p53 allele.Many ovarian tumors have the type of mutation withaggregation-driven dominant-negative effect.3,11 Thesemutations can inactivate wt p53 allele by aggregating with wtp53 protein. A number of these mutations also exert gain offunctions, such as binding to new target DNA or proteins. Thus,the behaviors of mutant p53 aggregates are more complex anddifficult to study. Our model provides the basis for thecharacterization on how mutant p53 aggregates affect p53function and chemoresponse. The formation of aggregates incancer cells is neither limited to p53 protein. Our staining resultsshowed that the anti-amyloid fibril antibody staining did notcompletely overlap with p53 staining in the cytoplasm, indicatingthat there are other proteins involved into aggregation in thecytoplasm. When p53 was very effectively knocked down byshRNA, we detected significant decreased levels of aggregates byanti-amyloid fibril dot blot (Figure 4b). However, low levels ofaggregates were still present. The staining result showed that

knockdown of p53 diminished the nuclear staining of anti-amyloid fibril antibody (Supplementary Figure 7), but low levelsof aggregates were still present in the cytoplasm. These resultsalso indicate the existence of aggregates formed by otherproteins in the cytoplasm.The association of p53 aggregation and tumor progression was

recently reported.11,46,47 For instance, by coaggregation, p53aggregation can inactivate p63 and p73, leading to an increase inoncogenic potential in cells.11 The mechanisms of p53 aggrega-tion promoting tumor progression still need to be furtherelucidated. In our system, the protein-binding profile of AGp53reveals potential related pathways (Table 1). Cytoskeleton proteinsare among AGp53-associated proteins. This is consistent with aprevious report showing that protein aggregates interact withcytoskeletal architecture,20 thus interfering with intracellulartransport and cell division.We found a group of translation initiation factors (EIF2 and EIF3

subunits) interacting with AGp53, possibly altering global proteinsynthesis of OCSCs. Notably, AGp53 shows increased binding withtopoisomerase-1 (TOP1) that may cause increased DNA repair andgenetic instability in OCSCs, as p53 enhances TOP1 catalyticactivities of genetic recombination.48 In addition, AGp53 interac-tion with heat shock protein 90 (HSP90) strengthens our theorythat protein aggregation drives wt p53 into the mutantphenotypic form, as only the mutant p53 recruits HSP90 to inhibitubiquitination turnover, resulting in p53 stabilization.49 Studyingthe protumor cellular functions of AGp53 will identify noveltherapeutic targets.Overexpression of p53 in tumors is generally interpreted as a

sign of TP53 gene mutation and represents a hallmark of cancer.However, the independent prognostic value of aberrant p53 status(expression level and mutation) remains controversial. Theaggregation of p53 is a new parameter that helps us assess p53status. Using fibril-specific, conformation-dependent antibodiesrecognizing generic epitopes, protein aggregates can be identifiedwithin tumor samples. By co-staining with p53 it is possible todetect p53 aggregates in tumor biopsies as a marker for platinumresistance. As many of the HGSOC tumors retain at least one wtp53 allele, reactivation of wt p53 through suppressing aggregationis a promising therapeutic strategy.Our study demonstrates a cell-specific regulation of p53 protein

stabilization associated with protein aggregation that conferschemoresistance. The correlation between p53 aggregation andplatinum resistance is relevant to understand the poor prognosisof patients with HGSOC. Use of p53 aggregation as a markerwould allow the stratification of these patients and personalizedtherapy. Furthermore, aggregation inhibitors can be potentialchemosensitizers in combination with platinum-based therapiesto eliminate resistant cancer stem cells and prevent recurrence(Figure 6). In order to fully understand the role of p53 aggregationin ovarian cancer and its value as a prognostic marker or atherapeutic target, cancer cell lines and tumor samples with a

Figure 3. The overexpression of ARF causes p53 accumulation in OCSCs. (a) Western blots of p53 and MDM2 in OCSC1s and OCC1s treated bycarboplatin. (b) Reverse transcriptase–quantitative PCR (RT–qPCR) analysis of MDM2 expression in OCSC1s and OCC1s. (c) Western blots ofMDM2 in OCSC1s treated by 5 μM MG132 for 24 h. (d) Western blots of MDM2 and p53 in OCSC1s and OCC1s transfected by pcDNA-MDM2 (+)or control vector (− ) plasmid. (e) RT–qPCR analysis (upper) of ARF mRNA and western blots (lower) of ARF, MDM2 and p53 in control pLKO-OCSCs (infected by pLKO vector), shARF1-OCSCs and shARF2-OCSCs (infected by lentiviral shRNAs targeting ARF). (f) Gel electrophoresis ofmethylation-specific PCR (MS-PCR). DNA products were amplified from bisulfite-converted DNA of OCSCs and OCCs. The unmethylated (U)and methylated (M) ARF promoters were detected by specific primers. Normal ovarian surface epithelial (OSE) cell DNA is used as a negativecontrol. CpG Methyltransferase (M.SssI)-treated OSE DNA is a positive control for ARF promoter methylation. (g) MSPCR analysis of ARFpromoter (M, methylated; U, unmethylated) and RT–qPCR analysis of CD44 expression in cancer cell lines derived from 17 advanced ovariantumor biopsies. They were classified as OCSC type (n= 10) or OCC type (n= 7) based on their phenotypes. Data are normalized to OCSC1s. Theabove data represent the means± s.d. of three technical replicates. (h) RT–qPCR analysis of ARF expression in OCSC-type and OCC-type cancercell lines. Data are normalized to OCSC1s. Each dot represents one cell line. Bars represent the mean of each group. The two groups aresignificantly different (P-value is indicated in the graph).


7


Figure 4. ARF silencing reverses p53 aggregation and sensitizes OCSCs in vitro. (a) Schematic model of p53 regulation in OCSCs (left) and OCCs(right). (b) Dot-blot analysis of amyloid fibrils in control pLKO-OCSC1s, shARF1-OCSC1s, shARF2-OCSC1s and shp53-OCSC1s. (c) Viability ofshARF1-OCSC1s and shARF2-OCSC1s treated by carboplatin. pLKO-OCSC1s and OCC1s are included as chemo-resistant and -sensitivecontrols, respectively. Data were normalized to respective no treatment controls (NT). *Po0.05 vs NT. (d) Fluorescence images of co-culturemodel mimicking heterogeneous tumors. GFP-labeled OCSCs or GFP-labeled shARF2-OCSC1s (green) and mCherry-labeled OCCs (red) weremixed and co-cultured. They were treated by carboplatin for 48 h followed by recovery for 48 h. Scale bars are 36 μm.


8


Figure 5. ARF silencing reverses p53 aggregation and sensitizes OCSCs in vivo. (a) Representative images of cisplatin-treated mice with tumors(red) formed by mCherry-labeled pLKO-OCSC1s or shARF2-OCSC1s. (b) Growth curves of cisplatin-treated tumors formed by control pLKO-OCSC1s (left, n= 5) and shARF2-OCSC1s (right, n= 5). Tumor size was quantified by the total number of pixels in the fluorescent areas of theimage. Each curve represents the tumor growth in one mouse. The red dotted line indicates the cutoff baseline between stable disease andprogression. (c) P53 western blot of non-denaturing gel analysis of tumors formed by pLKO-OCSC1s (lanes 1 and 2) or shARF2-OCSC1s (lanes 3and 4) nontreated (− ) or treated (+) with cisplatin. The aggregated p53 protein was detected as the p53 bands at high molecular weight (highMW p53). (d, e) Caspase-3 and caspase-9 activity in cisplatin-treated tumors (n= 3; ✶Po0.005) formed by pLKO-OCSC1s (pLKO) and shARF-OCSC1s (shARF). (f) Hematoxylin and eosin (H&E) staining of tumors from mice treated by cisplatin.


9


wide spectrum of p53 mutations will be further investigated in ourfuture study.

MATERIALS AND METHODSHuman tissue specimens and cellsCD44+ ovarian cancer cells were isolated from stage III/IV serous ovariancarcinoma samples as previously described.11 Sample collection was

carried out with patient consent and approved by the Human Investiga-tions Committee of Yale University School of Medicine. Ovarian cancer cellswere propagated as previously described.11,12

Flow cytometryFlow cytometry analysis was performed as previously described.29

Antibodies include anti-CD44-FITCe (no. 11-0441-81, eBioscience, SanDiego, CA, USA) and control Isotype-FITC (Sungene Biotech, Tianjin, China).

Transient transfectionCells were transfected with pcDNA3-MDM2 (Addgene, Cambridge, MA,USA, plasmid no. 16233) or pcDNA3 (Invitrogen, Carlsbad, CA, USA)plasmid using FuGENE 6 Transfection Reagent (Roche Diagnostics,Indianapolis, IN, USA) according to the manufacturer’s instructions.Transfected cells were grown up to 48 h before proteins were extracted.

Lentiviral infectionsTwo lentiviral shRNAs targeting ARF and pLKO.1-puro empty vector (OpenBiosystems, Huntsville, AL, USA) were produced as previously described.50

The shRNA and control lentivirus were introduced to OCSC cells by addingto cell growth medium. Then, stable knockdown or control cells wereselected by medium containing 1 μg/ml puromycin (Sigma-Aldrich,Milwaukee, WI, USA).

Cell growth and viability assaysCell growth and morphology were assessed using Incucyte (EssenInstruments, Ann Arbor, MN, USA). Cell viability was determined withCellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega,Madison, WI, USA) according to the manufacturer’s instructions. Theviability of treated cells was normalized to the untreated control.

ImmunoblottingWestern blot was performed as previously described.51 Antibodies werediluted as following: anti-p14ARF (1:1000, no. 2407; Cell SignalingTechnology, Beverly, MA, USA), anti-p53 (1:2000, no. OP43; Millipore,Billerica, MA, USA), anti-MDM2 (1:1000, no. OP115; Millipore) and GAPDH(1:10 000, Sigma-Aldrich).Dot-blot analysis was performed as previously described.52

Anti-amyloid fibril OC (1:1000, no. AB2286; Millipore) was used todetect protein aggregates. The intensity of the dots was analyzed

Figure 6. Schematic model of chemosensitization targeting p53 protein aggregation.

Table 1. The list of proteins binding to aggregated p53 protein inOCSCs

Cellular functions P53 aggregates binding proteins

Cytoskeleton-relatedproteins

PLEC1, MYH9, MYH10, EPPK1, ACTN1,AXTN4, DYNC1H1, FLNA, FLNC, SPTAN1,

SPTBN1, EVPL, VIMMolecular chaperones HSP90AA1,HSP90AB1, HSP90AA2, HSP90B1Scaffold proteins IQGAP1Exocytosis ANXA1Cytokinesis CITProtein degradation PSMD2Translation EIF2A, EIF2S1, EIF3A, EIF3B, EIF3J, EEF1D,

RPS3tRNA synthetases QARSMitochondria IMMT, NDUFA9Metabolism FASN, CAD, MOGS, MTHFD1, LDHBNuclear structure AHNAKRibosome biogenesis RPLP0, RPL5, RPL6RNA processing DDX21, SFPQ, CDC5L, WBP11, ILF3Transcription NCL, SFPQ, SND1, CDC5LDNA replication MCM3, MCM4, MCM5, TOP1Unspecified EPPK1, TTC13, NPEPPS

Abbreviations: OCSC, ovarian cancer stem cell; tRNA, transfer RNA. OCSC1sand OCC1s were treated with carboplatin. Their p53 protein complex wasimmunoprecipitated and analyzed by two-dimensional gel electrophoresis.The proteins that showed significantly greater (45-fold) enrichment in p53complex of OCSC1s were identified by mass spectrometry. They are listedand classified according to their cellular functions.


10


using Image Analysis Software (Kodak Scientific Imaging Systems,Rochester, NY, USA).

Quantitative real-time PCRTotal RNA was extracted from cells or tumor tissues using an RNeasy MiniKit (Qiagen, Austin, TX, USA) according to the manufacturer’s instructions.cDNA was synthesized with Verso cDNA Kit (Thermo Scientific, Waltham,MA, USA) according to the manufacturer’s instructions. Quantification ofmRNA was performed using KAPA SYBR FAST qPCR Kit (KAPA Biosystems,Woburn, MA, USA) and CFX96TM PCR detection system (Bio-Rad, Hercules,CA, USA). GAPDH was used as reference gene. Relative expression wascalculated using the comparative ΔΔCT method. Primer sequences arelisted in Supplementary Table 4.

P53 cDNA and DNA sequencingThe p53 cDNA was used as a template to amplified full-length p53 by PCRreaction as previously described.53 Genomic DNA was extracted with aDNeasy Blood and Tissue Kit (Qiagen). Exons 5–8 of TP53 were amplified byPCR. The purified PCR products were sequenced using the same primers asPCR listed in Supplementary Table 4.

Chromatin immunoprecipitationChromatin immunoprecipitation was performed as previously described.54

Anti-p53 antibody (no. OP43, EMD Millipore, Temecula, CA, USA) wasadded to the test groups. Mouse IgG (sc-2025, Santa Cruz Biotechnology,Dallas, TX, USA) was added to the immunoprecipitation (IP) control groups.Quantitative PCR primer sequences are listed in Supplementary Table 4.

GAPDH is used as the negative control. The PCR data of IP test and IgGcontrol groups were both normalized to input control group. Afternormalization, the signals of IP p53 samples are divided by the signals ofIgG control sample to get the fold of increase in IP p53 signal relative tothe background signal. At last, the data of OCSC1 and OCC1 were dividedby the nontreated OCSC1. The data of OCSC2 and OCC2 were divided bythe nontreated OCC2. The final step is comparing p53 DNA bindingbetween OCSCs and their daughter cell lines OCCs upon carboplatintreatment.

Methylation-specific PCRDNA methylation patterns in the CpG islands of the p14ARF gene weredetermined by methylation-specific PCR as previously described.55 Primersequences are listed in Supplementary Table 4. Normal human ovariansurface epithelial cells were used as negative control (unmethylatedalleles). The DNA of ovarian surface epithelial cells that was treated in vitrowith SssI CpG methyltransferase at 37 °C for 1 h was used as a positivecontrol (methylated alleles).

Immunocytofluorescence costainingAdherent cells on 8-chamber glass LabTek culture slides (BD Falcon,Bedford, MA, USA) were fixed with ice-cold 3.7% paraformaldehyde for5 min and permeabilized using 0.2% Triton X-100. The slides were thenblocked in 10% normal goat serum and incubated at room temperaturewith anti-p53 antibody (1:500, no. OP43, EMD Millipore) for 1 h and thenAlexaFluor-546-labeled goat anti-mouse IgG antibody (Invitrogen) for 1 h.For amyloid fibril staining, slides were incubated with anti-amyloid fibrilantibody (1:1000, no. AB2286 Millipore) at 4 °C overnight, then withAlexaFluor-488-labeled donkey anti-rabbit IgG antibody (1:1000, LifeTechnologies, Carlsbad, CA, USA) at room temperature for 1 h. Forthioflavin T staining, after p53 staining, slides were incubated in thioflavinT (5 mM in phosphate-buffered saline) for 10min at room temperature. Theslides were washed with 70% ethanol for 5 min twice and then rinsed withdistilled water 3 times.Nuclei were stained using 4′,6-diamidino-2-phenylindole. Between

each blocking or staining step, the slides were washed by phosphate-buffered saline 3 times. The slides were cover-slipped using FluorescenceMounting Medium (Dako, Glostrup, Denmark) and captured with ZeissAxio Observer inverted fluorescence microscope (Carl Zeiss, Oberkochen,Germany). Acquisition settings were kept constant.

Identification of aggregated p53-binding proteinsImmunoprecipitation of p53 was performed on carboplatin (100 μg/ml,24 h)-treated OCSC1s and OCC1s as previously described56 withanti-p53 (1 μg/ml, no. OP43, EMD Millipore) antibody. The precipitatedp53 complex was proceeded to two-dimensional gel. The two-dimensional gel analysis was performed as previously described.57

The gel spots with fivefold greater intensity in the gel of OCSC1sthan OCC1s were picked for identification by liquid chromatographytandem mass spectrometry) analysis. The liquid chromatographytandem mass spectrometry was performed as previously described.58

Proteins that had at least three fragments identified by liquidchromatography tandem mass spectrometry were selected as positivecandidates.

Xenograft experimentsYale University Institutional Animal Care and Use Committee approved thisprotocol. Six-week-old female athymic nude mice (Harlan, South Easton, MA,USA) were maintained under pathogen-free conditions, and food and waterwere supplied ad libitum. In the subcutaneous and intraperitoneal xenograftmodels, 3 × 106 and 7×106 cells were injected, respectively. At 7 days postinjection, mice were treated with cisplatin at 5 μg/kg body weight once perweek. Tumors were monitored by Carestream In-Vivo Imaging System.Images were quantified by Carestream software (Carestream Health,Woodbridge, CT, USA). Tumor size was measured by creating an automaticregion of interest and measuring that area’s pixel number.

Statistical analysesNumerical values are reported as the mean ± s.d. For comparisonsbetween two groups, P-values were calculated using unpaired two-tailed Student’s t-tests. For the correlation analysis, P-values werecalculated using Fisher’s exact test. P-values of ⩽ 0.05 were consideredstatistically significant.

CONFLICT OF INTERESTThe authors declare no conflict of interest.

ACKNOWLEDGEMENTSThis study was supported by NIH Grants RO1CA118678 and RO1CA127913, SandsFamily Foundation and Discovery to Cure Program. We thank Drs Alice Soragni andDavid Eisenberg of UCLA for their suggestion that aggregated p53 could act as apotential therapeutic target.

REFERENCES1 Clarke-Pearson DL. Clinical practice. Screening for ovarian cancer. N Engl J Med

2009; 361: 170–177.2 Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian

carcinoma. Nature 2011; 474: 609–615.3 UMD TP53 mutation database. http://p53.fr/.4 IARC TP53 Database. http://www.iarc.fr/p53/Index.html.5 Freed-Pastor WA, Prives C. Mutant p53: one name, many proteins. Genes Dev

2012; 26: 1268–1286.6 Moll UM, Riou G, Levine AJ. Two distinct mechanisms alter p53 in breast

cancer: mutation and nuclear exclusion. Proc Natl Acad Sci USA 1992; 89:7262–7266.

7 Tominaga O, Hamelin R, Trouvat V, Salmon RJ, Lesec G, Thomas G et al. Frequentlyelevated content of immunochemically defined wild-type p53 protein in color-ectal adenomas. Oncogene 1993; 8: 2653–2658.

8 Houben R, Hesbacher S, Schmid CP, Kauczok CS, Flohr U, Haferkamp S et al.High-level expression of wild-type p53 in melanoma cells is frequentlyassociated with inactivity in p53 reporter gene assays. PLoS ONE 2011; 6:e22096.

9 Wong KK, Izaguirre DI, Kwan SY, King ER, Deavers MT, Sood AK et al. Poor survivalwith wild-type TP53 ovarian cancer? Gynecol Oncol 2013; 130: 565–569.

10 Silva JL, Rangel LP, Costa DC, Cordeiro Y, De Moura Gallo CV. Expanding the prionconcept to cancer biology: dominant-negative effect of aggregates of mutant p53tumour suppressor. Biosci Rep 2013; 33: 593–603.

11 Xu J, Reumers J, Couceiro JR, De Smet F, Gallardo R, Rudyak S et al. Gain offunction of mutant p53 by coaggregation with multiple tumor suppressors. NatChem Biol 2011; 7: 285–295.


11


12 Ano Bom AP, Rangel LP, Costa DC, de Oliveira GA, Sanches D, Braga CA et al.Mutant p53 aggregates into prion-like amyloid oligomers and fibrils: implicationsfor cancer. J Biol Chem 2012; 287: 28152–28162.

13 Levy CB, Stumbo AC, Ano Bom AP, Portari EA, Cordeiro Y, Silva JL et al. Co-localization of mutant p53 and amyloid-like protein aggregates in breast tumors.Int J Biochem Cell Biol 2011; 43: 60–64.

14 Stefani M, Dobson CM. Protein aggregation and aggregate toxicity: new insightsinto protein folding, misfolding diseases and biological evolution. J Mol Med (Berl)2003; 81: 678–699.

15 Higashimoto Y, Asanomi Y, Takakusagi S, Lewis MS, Uosaki K, Durell SR et al.Unfolding, aggregation, and amyloid formation by the tetramerizationdomain from mutant p53 associated with lung cancer. Biochemistry 2006; 45:1608–1619.

16 Ishimaru D, Andrade LR, Teixeira LS, Quesado PA, Maiolino LM, Lopez PM et al.Fibrillar aggregates of the tumor suppressor p53 core domain. Biochemistry 2003;42: 9022–9027.

17 Rigacci S, Bucciantini M, Relini A, Pesce A, Gliozzi A, Berti A et al. The (1-63) regionof the p53 transactivation domain aggregates in vitro into cytotoxic amyloidassemblies. Biophys J 2008; 94: 3635–3646.

18 Butler JS, Loh SN. Structure, function, and aggregation of the zinc-free form of thep53 DNA binding domain. Biochemistry 2003; 42: 2396–2403.

19 Ishimaru D, Lima LM, Maia LF, Lopez PM, Ano Bom AP, Valente AP et al. Reversibleaggregation plays a crucial role on the folding landscape of p53 core domain.Biophys J 2004; 87: 2691–2700.

20 Olzscha H, Schermann SM, Woerner AC, Pinkert S, Hecht MH, Tartaglia GG et al.Amyloid-like aggregates sequester numerous metastable proteins with essentialcellular functions. Cell 2011; 144: 67–78.

21 Forget KJ, Tremblay G, Roucou X p53 aggregates penetrate cells and induce theco-aggregation of intracellular p53. PLoS ONE 2013; 8: e69242.

22 Lasagna-Reeves CA, Clos AL, Castillo-Carranza D, Sengupta U,Guerrero-Munoz M, Kelly B et al. Dual role of p53 amyloid formation in cancer;loss of function and gain of toxicity. Biochem Biophys Res Commun 2013; 430:963–968.

23 Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF. Prospectiveidentification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 2003; 100:3983–3988.

24 Richardson GD, Robson CN, Lang SH, Neal DE, Maitland NJ, Collins AT. CD133, anovel marker for human prostatic epithelial stem cells. J Cell Sci 2004; 117(Pt 16):3539–3545.

25 Shih Ie M, Davidson B. Pathogenesis of ovarian cancer: clues from selectedoverexpressed genes. Future Oncol 2009; 5: 1641–1657.

26 Peng S, Maihle NJ, Huang Y. Pluripotency factors Lin28 and Oct4 identify a sub-population of stem cell-like cells in ovarian cancer. Oncogene 2010; 29:2153–2159.

27 Rizzo S, Hersey JM, Mellor P, Dai W, Santos-Silva A, Liber D et al. Ovarian cancerstem cell-like side populations are enriched following chemotherapy andoverexpress EZH2. Mol Cancer Ther 2011; 10: 325–335.

28 Steg AD, Bevis KS, Katre AA, Ziebarth A, Dobbin ZC, Alvarez RD et al. Stem cellpathways contribute to clinical chemoresistance in ovarian cancer. Clin Cancer Res2012; 18: 869–881.

29 Alvero AB, Chen R, Fu HH, Montagna M, Schwartz PE, Rutherford T et al.Molecular phenotyping of human ovarian cancer stem cellsunravels the mechanisms for repair and chemoresistance. Cell Cycle 2009; 8:158–166.

30 Alvero AB, Fu HH, Holmberg J, Visintin I, Mor L, Marquina CC et al. Stem-likeovarian cancer cells can serve as tumor vascular progenitors. Stem Cells 2009; 27:2405–2413.

31 Chen R, Alvero AB, Silasi DA, Kelly MG, Fest S, Visintin I et al. Regulation of IKKbetaby miR-199a affects NF-kappaB activity in ovarian cancer cells. Oncogene 2008; 27:4712–4723.

32 Craveiro V, Yang-Hartwich Y, Holmberg JC, Sumi NJ, Pizzonia J, Griffin B et al.Phenotypic modifications in ovarian cancer stem cells following Paclitaxel treat-ment. Cancer Med 2013; 2: 751–762.

33 Aguzzi A, O'Connor T. Protein aggregation diseases: pathogenicity and ther-apeutic perspectives. Nat Rev Drug Discov 2010; 9: 237–248.

34 Zhang Y, Xiong Y, Yarbrough WG. ARF promotes MDM2 degradation and stabi-lizes p53: ARF-INK4a locus deletion impairs both the Rb and p53 tumor sup-pression pathways. Cell 1998; 92: 725–734.

35 Pomerantz J, Schreiber-Agus N, Liegeois NJ, Silverman A, Alland L, Chin L et al.The Ink4a tumor suppressor gene product, p19Arf, interacts with MDM2 andneutralizes MDM2's inhibition of p53. Cell 1998; 92: 713–723.

36 Tao W, Levine AJ. P19(ARF) stabilizes p53 by blocking nucleo-cytoplasmic shut-tling of Mdm2. Proc Natl Acad Sci USA 1999; 96: 6937–6941.

37 Weber JD, Taylor LJ, Roussel MF, Sherr CJ, Bar-Sagi D. Nucleolar Arf sequestersMdm2 and activates p53. Nat Cell Biol 1999; 1: 20–26.

38 Iwakuma T, Lozano G MDM2 an introduction. Mol Cancer Res 2003; 1: 993–1000.39 Rayburn E, Zhang R, He J, Wang H. MDM2 and human malignancies: expression,

clinical pathology, prognostic markers, and implications for chemotherapy. CurrCancer Drug Targets 2005; 5: 27–41.

40 Biancalana M, Koide S. Molecular mechanism of Thioflavin-T binding to amyloidfibrils. Biochim Biophys Acta 2010; 1804: 1405–1412.

41 Kayed R, Head E, Sarsoza F, Saing T, Cotman CW, Necula M et al. Fibril specific,conformation dependent antibodies recognize a generic epitope common toamyloid fibrils and fibrillar oligomers that is absent in prefibrillar oligomers. MolNeurodegener 2007; 2: 18.

42 Zheng S, Chen P, McMillan A, Lafuente A, Lafuente MJ, Ballesta A et al.Correlations of partial and extensive methylation at the p14(ARF)locus with reduced mRNA expression in colorectal cancer cell linesand clinicopathological features in primary tumors. Carcinogenesis 2000; 21:2057–2064.

43 Ciechanover A. The ubiquitin proteolytic system: from a vague idea, through basicmechanisms, and onto human diseases and drug targeting. Neurology 2006;66(2 Suppl 1): S7–19.

44 Wong E, Cuervo AM. Autophagy gone awry in neurodegenerative diseases. NatNeurosci 2010; 13: 805–811.

45 Kraiss S, Spiess S, Reihsaus E, Montenarh M. Correlation of metabolic stability andaltered quaternary structure of oncoprotein p53 with cell transformation. Exp CellRes 1991; 192: 157–164.

46 Wilcken R, Wang G, Boeckler FM, Fersht AR. Kinetic mechanism of p53 oncogenicmutant aggregation and its inhibition. Proc Natl Acad Sci USA 2012; 109:13584–13589.

47 Silva JL, Gallo CV, Costa DC, Rangel LP. Prion-like aggregation of mutant p53in cancer. Trends Biochem Sci 2014; 39: 260–267.

48 Albor A, Kaku S, Kulesz-Martin M. Wild-type and mutant forms of p53 activatehuman topoisomerase I: a possible mechanism for gain of function in mutants.Cancer Res 1998; 58: 2091–2094.

49 Peng Y, Chen L, Li C, Lu W, Chen J. Inhibition of MDM2 by hsp90 contributes tomutant p53 stabilization. J Biol Chem 2001; 276: 40583–40590.

50 Tiscornia G, Singer O, Verma IM. Production and purification of lentiviral vectors.Nat Protoc 2006; 1: 241–245.

51 Kamsteeg M, Rutherford T, Sapi E, Hanczaruk B, Shahabi S, Flick M et al. Phe-noxodiol--an isoflavone analog--induces apoptosis in chemoresistant ovariancancer cells. Oncogene 2003; 22: 2611–2620.

52 Kayed R, Glabe CG. Conformation-dependent anti-amyloid oligomer antibodies.Methods Enzymol 2006; 413: 326–344.

53 Szybka M, Zakrzewska M, Rieske P, Pasz-Walczak G, Kulczycka-Wojdala D, Zawlik Iet al. cDNA sequencing improves the detection of P53 missense mutations incolorectal cancer. BMC Cancer 2009; 9: 278.

54 Wells J, Farnham PJ. Characterizing transcription factor binding sitesusing formaldehyde crosslinking and immunoprecipitation. Methods 2002; 26:48–56.

55 Klump B, Hsieh CJ, Dette S, Holzmann K, Kiebetalich R, Jung M et al. Promotermethylation of INK4a/ARF as detected in bile-significance for the differentialdiagnosis in biliary disease. Clin Cancer Res 2003; 9: 1773–1778.

56 Erica Golemis PDA (ed) Protein-Protein Interactions: A Molecular Cloning Manual.Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, 2002.

57 Wu TL. Two-dimensional difference gel electrophoresis. Methods Mol Biol 2006;328: 71–95.

58 Dieguez-Acuna F, Kodama S, Okubo Y, Paz AC, Gygi SP, Faustman DL. Proteomicsidentifies multipotent and low oncogenic risk stem cells of the spleen. Int JBiochem Cell Biol 2010; 42: 1651–1660.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)


12


Documents

p53 protein aggregation promotes platinum resistance in ......ORIGINAL ARTICLE p53 protein aggregation promotes platinum resistance in ovarian cancer Y Yang-Hartwich 1, MG Soteras